Fingerprint Presentation Attack Detection by Learning in Frequency Domain

Abstract

The anti-spoofing ability is of great importance to the development of automated fingerprint recognition systems. This paper proposes a novel Optical Coherence Technology (OCT)-based fingerprint Presentation Attack Detection (PAD) method from frequency domain. Unlike previous approaches, we design an Frequency Feature Disentangling (FFD) model to decompose OCT-based fingerprint B-scans into four different frequency subbands like Discrete Wavelet Transform (DWT). Through such disentangling, information superimposed in original image in spatial domain (e.g. discriminative PAD feature, invalid and redundant feature) can be separated respectively. We then let it learn different frequency codes to form their corresponding latent codes. Spoofness score which is used to distinguish PAs from bonafides is finally designed based on the latent codes. The experimental results, evaluated on the dataset with 93,200 bonafide B-scans from 137 fingers and 48,400 B-scans from 121 PAs, show that our method can remove some useless interference information superimposed in spatial domain by disentangling into frequency domain for effective PAD. In the instance-wise case, the proposed method achieves a minimum error (Err.) of 0.67%, which outperforms other compared methods and improves 81.89% than the best one.